@@ -45,9 +45,8 @@ const fileParsing = async (options: any, req: any) => {
4545 return options . messages
4646}
4747
48- openaiRouter . post ( '/stream/:version? ' , upload . single ( 'file' ) , async ( r , res ) => {
48+ openaiRouter . post ( '/stream/v2 ' , upload . single ( 'file' ) , async ( r , res ) => {
4949 const req = r as RequestWithUser
50- const { version } = r . params
5150 const { options, courseId } = JSON . parse ( req . body . data )
5251 const { model, userConsent } = options
5352 const { user } = req
@@ -105,31 +104,134 @@ openaiRouter.post('/stream/:version?', upload.single('file'), async (r, res) =>
105104 model,
106105 } )
107106
108- let events
109- if ( version === 'v2' ) {
110- const latestMessage = options . messages [ options . messages . length - 1 ] // Adhoc to input only the latest message
111- events = await responsesClient . createResponse ( { input : [ latestMessage ] , prevResponseId : options . prevResponseId } )
112- } else {
113- events = await getCompletionEvents ( options )
114- }
107+ const latestMessage = options . messages [ options . messages . length - 1 ] // Adhoc to input only the latest message
108+ const events = await responsesClient . createResponse ( { input : [ latestMessage ] , prevResponseId : options . prevResponseId } )
109+
115110 if ( isError ( events ) ) {
116111 res . status ( 424 )
117112 return
118113 }
119114
120115 res . setHeader ( 'content-type' , 'text/event-stream' )
121116
122- let completion
123- if ( version === 'v2' ) {
124- completion = await responsesClient . handleResponse ( {
125- events,
126- encoding,
127- res,
117+ const completion = await responsesClient . handleResponse ( {
118+ events,
119+ encoding,
120+ res,
121+ } )
122+
123+ tokenCount += completion . tokenCount
124+
125+ let userToCharge = user
126+ if ( inProduction && req . hijackedBy ) {
127+ userToCharge = req . hijackedBy
128+ }
129+
130+ if ( courseId ) {
131+ await incrementCourseUsage ( userToCharge , courseId , tokenCount )
132+ } else if ( model !== FREE_MODEL ) {
133+ await incrementUsage ( userToCharge , tokenCount )
134+ }
135+
136+ logger . info ( `Stream ended. Total tokens: ${ tokenCount } , {
137+ tokenCount,
138+ model,
139+ user : user . username ,
140+ courseId,
141+ } )
142+
143+ const course =
144+ courseId &&
145+ ( await ChatInstance . findOne ( {
146+ where : { courseId } ,
147+ } ) )
148+
149+ const consentToSave = courseId && course . saveDiscussions && options . saveConsent
150+
151+ console . log ( 'consentToSave' , options . saveConsent , user . username )
152+
153+ if ( consentToSave ) {
154+ const discussion = {
155+ userId : user . id ,
156+ courseId,
157+ response : completion . response ,
158+ metadata : options ,
159+ }
160+ await Discussion . create ( discussion )
161+ }
162+
163+ encoding . free ( )
164+
165+ res . end ( )
166+ return
167+ } )
168+
169+ openaiRouter . post ( '/stream' , upload . single ( 'file' ) , async ( r , res ) => {
170+ const req = r as RequestWithUser
171+ const { options, courseId } = JSON . parse ( req . body . data )
172+ const { model, userConsent } = options
173+ const { user } = req
174+
175+ options . options = { temperature : options . modelTemperature }
176+
177+ if ( ! user . id ) {
178+ res . status ( 401 ) . send ( 'Unauthorized' )
179+ return
180+ }
181+
182+ const usageAllowed = courseId ? await checkCourseUsage ( user , courseId ) : model === FREE_MODEL || ( await checkUsage ( user , model ) )
183+
184+ if ( ! usageAllowed ) {
185+ res . status ( 403 ) . send ( 'Usage limit reached' )
186+ return
187+ }
188+
189+ let optionsMessagesWithFile = null
190+
191+ try {
192+ if ( req . file ) {
193+ optionsMessagesWithFile = await fileParsing ( options , req )
194+ }
195+ } catch ( error ) {
196+ logger . error ( 'Error parsing file' , { error } )
197+ res . status ( 400 ) . send ( 'Error parsing file' )
198+ return
199+ }
200+
201+ options . messages = getMessageContext ( optionsMessagesWithFile || options . messages )
202+ options . stream = true
203+
204+ const encoding = getEncoding ( model )
205+ let tokenCount = calculateUsage ( options , encoding )
206+ const tokenUsagePercentage = Math . round ( ( tokenCount / DEFAULT_TOKEN_LIMIT ) * 100 )
207+
208+ if ( model !== FREE_MODEL && tokenCount > 0.1 * DEFAULT_TOKEN_LIMIT && ! userConsent ) {
209+ res . status ( 201 ) . json ( {
210+ tokenConsumtionWarning : true ,
211+ message : `You are about to use ${ tokenUsagePercentage } ,
128212 } )
129- } else {
130- completion = await streamCompletion ( events , options , encoding , res )
213+ return
131214 }
132215
216+ const contextLimit = getModelContextLimit ( model )
217+
218+ if ( tokenCount > contextLimit ) {
219+ logger . info ( 'Maximum context reached' )
220+ res . status ( 403 ) . send ( 'Model maximum context reached' )
221+ return
222+ }
223+
224+ const events = await getCompletionEvents ( options )
225+
226+ if ( isError ( events ) ) {
227+ res . status ( 424 )
228+ return
229+ }
230+
231+ res . setHeader ( 'content-type' , 'text/event-stream' )
232+
233+ const completion = await streamCompletion ( events , options , encoding , res )
234+
133235 tokenCount += completion . tokenCount
134236
135237 let userToCharge = user
0 commit comments